From a8471a3995ce5ff965179eab63e7e67906b797bd Mon Sep 17 00:00:00 2001 From: Tianyu Liu Date: Tue, 16 Apr 2024 11:00:16 -0700 Subject: [PATCH] polish toml files [ghstack-poisoned] --- train_configs/debug_model.toml | 4 ++-- train_configs/llama_13b.toml | 2 +- train_configs/llama_70b.toml | 2 +- train_configs/llama_7b.toml | 2 +- 4 files changed, 5 insertions(+), 5 deletions(-) diff --git a/train_configs/debug_model.toml b/train_configs/debug_model.toml index 6eb623a5e..baf6bd553 100644 --- a/train_configs/debug_model.toml +++ b/train_configs/debug_model.toml @@ -19,7 +19,7 @@ save_tb_folder = "tb" [model] name = "llama" flavor = "debugmodel" -norm_type = "fused_rmsnorm" # layernorm / np_layernorm / rmsnorm / fused_rmsnorm +norm_type = "fused_rmsnorm" # layernorm / np_layernorm / rmsnorm / fused_rmsnorm tokenizer_path = "./torchtitan/datasets/tokenizer/tokenizer.model" [optimizer] @@ -37,7 +37,7 @@ tensor_parallel_degree = 1 pipeline_parallel_degree = 1 fp8_linear = "" compile = false -dataset = "alpaca" # supported datasets: alpaca (52K), openwebtext (8M), c4 (177M) +dataset = "alpaca" # supported datasets: alpaca (52K), openwebtext (8M), c4 (177M) [checkpoint] enable_checkpoint = false diff --git a/train_configs/llama_13b.toml b/train_configs/llama_13b.toml index 4fc72c11c..7151df999 100644 --- a/train_configs/llama_13b.toml +++ b/train_configs/llama_13b.toml @@ -18,7 +18,7 @@ save_tb_folder = "tb" [model] name = "llama" flavor = "13B" -norm_type = "fused_rmsnorm" # [layernorm, np_layernorm, rmsnorm, fused_rmsnorm] +norm_type = "fused_rmsnorm" # [layernorm, np_layernorm, rmsnorm, fused_rmsnorm] tokenizer_path = "./torchtitan/datasets/tokenizer/tokenizer.model" [optimizer] diff --git a/train_configs/llama_70b.toml b/train_configs/llama_70b.toml index 1878647db..110453d4f 100644 --- a/train_configs/llama_70b.toml +++ b/train_configs/llama_70b.toml @@ -18,7 +18,7 @@ save_tb_folder = "tb" [model] name = "llama" flavor = "70B" -norm_type = "rmsnorm" # [layernorm, np_layernorm, rmsnorm, fused_rmsnorm] +norm_type = "rmsnorm" # [layernorm, np_layernorm, rmsnorm, fused_rmsnorm] tokenizer_path = "./torchtitan/datasets/tokenizer/tokenizer.model" [optimizer] diff --git a/train_configs/llama_7b.toml b/train_configs/llama_7b.toml index 7e8f7f789..3b8711517 100644 --- a/train_configs/llama_7b.toml +++ b/train_configs/llama_7b.toml @@ -17,7 +17,7 @@ save_tb_folder = "tb" [model] name = "llama" flavor = "7B" -norm_type = "fused_rmsnorm" # layernorm / np_layernorm / rmsnorm / fused_rmsnorm +norm_type = "fused_rmsnorm" # layernorm / np_layernorm / rmsnorm / fused_rmsnorm tokenizer_path = "./torchtitan/datasets/tokenizer/tokenizer.model" [optimizer]